JOINT ANALYSIS OF INFORMATIONAL SIGNS IN MULTI-PARAMETRIC COMBINED PATTERNS RECOGNITION SYSTEMS

Abstract

In modern science and technique is known a class of systems – multi-parametric patterns recognition systems. In such systems, recognition objects are represented by several patterns of different nature of origin. This allows to significantly increase the number of informative signs for the recognition object and increase the classification reliability. But in this case arised a negative aspect – increasing the time complexity of the data analyze processing. To eliminate this drawback is known the method of separate analysis of informative features, but it allows to obtain high accuracy of recognition only with a sufficiently large amount of data to be analyzed. This article considered question of the methods development for the joint analysis of information signs in multi-parametric combined pattern recognition systems. Joint analysis involves combining all the images that describe the object of recognition into one global image, the further selection of informative features and the decision on the basis of the minimum metrics. For minimizing the time complexity of data processing and decision making on classification, in article proposed to use methods for selecting informative signs. Using the informative signs selecting gives possibility to exclude form analysis less informative data and get a solution for fewer more informative parameters. The proposed technique using allows, in the case of a limited recognition object description and a small quantity of informative signs, to build an effective strategy for making a decision on classification, which provides increased reliability of recognition and reduced time complexity. The effectiveness of the proposed solution is confirmed from the standpoint of the statistical theory of recognition of Vapnik-Chervonenkis. The results are implemented in the information system for checking textual data for uniqueness.